Classification for a folded von Mises-Fisher distribution

When directional data fall on the positive orthant of the hypersphere, they can be modeled using a folded directional distribution. In this paper, we introduce the folded von Mises-Fisher distribution and propose the Bayes classification rule for this distribution. Then we evaluate the performance of this rule and we compare it with the classification rule for the von Mises-Fisher distribution. Finally, we present examples using spherical data from the literature, determining the error rate using the folded von Mises-Fisher rule and comparing it with the error rate obtained using the von Mises-Fisher rule.